作物模型是现实世界中生物和非生物因素相互作用的不完美近似，在某些情况下，模型的结构、输入和参数选择的不确定性可能超过模拟产量的时空变异性，从而限制了模型的可预测性。不确定性分析是基于模型的风险分析和决策的重要组成部分，它能够给风险分析人员和决策者提供模型输出结果的准确性程度。模型的不确定性有三种来源，即输入变量的不确定性、模型参数的不确定性和模型算法的不确定性。其中，模型输入变量的不确定性主要源于输入参数如气候、土壤等相对实际值的误差，这往往是由测量误差或者数据缺失所导致的。在模型应用时，由输入变量所导致的模型不确定性往往大于模型内部参数或算法所导致的不确定性。因此，分析作物生长模拟模型中各种输入变量对模型模拟结果的不确定性，可以为模型使用者提供模型应用的风险分析和决策支持，为提高模型输出结果的准确性提供一定的理论依据。 本研究以APSIM、CERES、Nwheat和WheatGrow小麦生长模拟模型为基础，通过设置不同的土壤和气候情景，探讨了土壤和气象输入变量所导致的模型模拟结果的不确定性。研究结果如下：(1)对于APSIM、CERES和Nwheat模型，S1.1.1.1、S1.2.2.2和S1.2.3.3土壤情景下的产量模拟结果与实际土壤情景（S1.2.3.4）下的模拟结果之间变异系数较小，表明在缺少深层土壤数据时，可采用上层的土壤数据来进行代替。对于WheatGrow模型而言，模拟结果的精度与土壤深层数据的关系较大，所以在应用时应尽可能获取详尽的土壤数据。此外，通过比较不同土壤特征参数对模型模拟结果的不确定性时发现，土壤萎蔫点、田间持水量、饱和含水量等与水分相关的土壤特性参数对模型结果的不确定性影响很大，而其他土壤特性参数对模型结果的不确定性影响则较小。另外，本研究还发现，气象情景对土壤层数和土壤参数导致的模型模拟结果不确定性的影响较小，表明对于模型中土壤输入参数的研究基本可以不考虑气候的影响。(2)在低温室气体排放情景下四个小麦生长模型产量模拟结果的不确定性均较小，而在高温室气体排放情景下产量模拟结果的不确定性较大。同一天内不同增温时间对模拟结果不确定性的影响则表现为，夜间增温>日间增温>全天增温。此外，在研究各气象指标对模型输出结果的不确定性时发现，各气象指标对模拟结果不确定性的大小随模型和研究区域的不同而改变。其中，北部麦区对模型结果不确定性影响最大的是降水，黄淮麦区和长江中下游麦区是降水和温度，西南麦区是辐射量。对APSIM和Nwheat模型模拟结果不确定性影响最大的是温度，对CERES模型是降水，而对WheatGrow模型则是辐射量和温度。(3)总的来说，气象模型和作物生长模型均会导致模型输出结果的不确定性，且由作物生长模型导致的不确定性要大于气象模型导致的不确定性。此外，在比较四个小麦生长模型土壤及气象输入参数对模型不确定性的影响发现，四个模型均表现为：土壤水分参数>气象指标>土壤层数>气候模式。表明在使用生长模型进行作物生产力评估时，应更注重生长模型的选择以及土壤水分数据和气象数据的精确程度。

Crop growth model is an imperfect approximation of the interaction between biotic and abiotic factors in the real world. In some cases, the uncertainty of the structure, input and parameter selection of the model may exceed the spatiotemporal variability of the simulated yield, thus limiting the model''''''''''''''''s predictability. Uncertainty analysis is an important part of model-based risk analysis and decision making. It can provide risk assessors and decision makers with the accuracy information of the model output. There are three sources of uncertainty in the crop growth model, namely, input variables, parameter values and model equations.Among these, the uncertainty of the model input variable is mainly due to the error of the input data, such as climate, soil and other relative actual values, which is often caused by measurement error or data loss. In the crop growth model application, the uncertainty of models caused by the input variables is often greater than the uncertainty caused by the internal parameters or equations. Therefore, analyzing the input variables of crop growth simulation model on the uncertainty of the model simulation results can provide the risk analysis and decision support for the model users, and provide a theoretical basis for improving the accuracy of the model output. In this study, the uncertainty of simulation results caused by soil and meteorological input variables in APSIM, CERES, Nwheat and WheatGrow crop models were analyzed based on the different soil and climate scenarios. The results are as follows: (1) For the APSIM, CERES and Nwheat models, the coefficient of the yield simulation results under S1.1.1.1, S1.2.2.2 , S1.2.3.3 soil scenarios and the actual soil scenarios (S1.2.3.4) is small, which indicates that the deeper soil data can be instead with the upper soil data when the soil data in deeper layers was absence.For the WheatGrow model, the accuracy of the simulation results is highly correlated with the depth of the soil, so detailed soil data should be obtained when using this model. In addition, by comparing the uncertainty of different soil characteristic parameters on the model simulation results, it is found that soil moisture parameters such as soil wilting point, field water holding capacity and saturated water content have great influence on the uncertainty of simulation results, while other soil property parameters have less impact on the uncertainty of the simulation results. Moreover, the study also found that meteorological scenarios have little effect on the uncertainty of model simulation results caused by soil layers and soil parameters, indicating that the study of soil input parameters in the model can basically ignore the impact of climate. (2) Under the low greenhouse gas emission scenario, the uncertainty of the yield simulation results of the four wheat growth models is small, while the uncertainty of the yield simulation results under the high greenhouse gas emission scenario is large. The effect of different warming time on the uncertainty of the simulation results in the same day is as follows: nighttime warming > daytime warming > all day warming. In addition, the uncertainty of the simulation results varies with the model and the study area. Among them, the precipitation has the greatest impact on the uncertainty of the model results in the Northern Subregion. In the Huang-Huai Subregion and the Middle-Lower Reaches of Yangzi River Subregion, the precipitation and temperature have the great effect on the uncertainty, and in the Southwest Subregion the radiation is the main uncertainty factor.Moreover, for APSIM and Nwheat model, temperature is the most important factor affecting the uncertainty of simulation results, precipitation is the main factor affecting the uncertainty of simulation results of CERES model, and radiation and temperature are the most important factors that affects the uncertainty of simulation results of WheatGrow model. (3) In general, both meteorological models and crop growth simulation models can lead to the uncertainty of simulation results, and the uncertainty caused by the crop growth model is greater than the uncertainty caused by the meteorological model. In addition, by comparing the effects of soiland meteorological input parameters on the simulation uncertaintyof the four wheat growth models, it is found that: soil moisture parameters > meteorological factors > soil layers > climate models. It is suggested that more attention should be paid to the selection of suitable growth simulation models and the accuracy of soil moisture data and meteorological data when applying the crop simiulation models to assess the crop productivity.